In the previous installment, I reviewed the headway patterns for the 501 Queen route outbound to terminals at Neville, Humber and Long Branch. Now I will turn to the operation of the west end of the route between Long Branch and Parkside Drive. I used Parkside (the east side of High Park, and the continuation of Keele Street south of Bloor) as the eastern end of the measurement because it is at the end of the private right-of-way and because this avoids problems with variations caused by operations at Roncesvalles.
The charts presented here show headways (the frequency and regularity of service) as well as link times (the time needed to traverse part of the route). Headways are important to riders because they show how predictable service will be, and they also bear on riding comfort because crowding is directly affected by regularity of service. Link times are important for service planning because they show where and when congestion occurs, and how predictable (or not) the running time between locations will be.
The data here come from the TTC’s vehicle monitoring system for the months of December and January 2008/09. (The process of converting these data into the charts presented here is described in the previous post.)
Weather was a major factor in transit operations over the past winter.
During December, for roughly the week before Christmas, there was a series of cold days and a considerable accumulation of snow that fouled traffic and transit operations. In January, the mean temperature stayed below freezing for the entire month and many days with snow, notably the 7th and 28th.
Link Times Eastbound
Pages 1 to 5 show data for weekdays in the five calendar weeks in December.
In week 1, the time needed to reach Parkside from Long Branch rises in the AM peak, then settles down through the day with a pronounced drop off in the evening. On Monday, December 1, the line was badly disrupted and appears to have been broken in two for much of the day. Vehicle tracking data is quite chaotic for this day. On Tuesday, December 2, there was a major delay west of Royal York at the end of the AM peak causing extended running times.
Week 2 shows slightly better-behaved values during the daytime, but more scatter in the evening late in the week. Week 3 is again quite predictable until Friday, Dec. 19 and the snowstorm of that weekend. Weeks 4 and 5 are the Christmas period and there are only small peak effects. Indeed, the early closing on Christmas eve shows up in a drop-off of link times starting around 3pm.
Page 6 shows a “cloud” of all of the weekday link times. The purpose of this is to see the overall scatter of values, and we can see that most of these lie in a fairly narrow region.
Pages 7 and 8, Saturdays and Sundays, contain some unusual trend lines. On December 6th, 13th and 14th, there was little or no streetcar service on the west end of the line and many buses operated instead. Data for these days are not reliable. I have left them in the charts to show how a change in route operations creates problems for vehicle monitoring. (Note that the swings off the page of some trend lines are caused by polynomial interpolation which does not work well with few data points.)
If we break down the line into segments, we can see the behaviour of each section.
Long Branch to Kipling Eastbound Link Times December 2008 show absolutely no peak effects, and the running times over this section are quite consistent throughout the month.
Kipling to Royal York Eastbound Link Times December 2008 show the same pattern.
Royal York to Humber Eastbound Link Times December 2008 show a small amount of peak period effect, but certainly nothing on the scale we see on routes downtown. Note that for this analysis, “Humber” is the Lake Shore side of Humber Loop used only by cars bound to and from Long Branch.
Overall, there is nothing in the inbound link times to suggest that any sort of right-of-way between Long Branch and Humber will make a noticeable difference in the running time.
Humber to Parkside Eastbound Link Times December 2008 are similarly well behaved, and this segment has its own right-of-way.
The small rises we see in the overall trip from Long Branch to Parkside are the accumulations of even smaller variations along the line, some of which are likely the result of loading times, not congestion.
Link Times Westbound
Parkside to Long Branch Westbound Link Times December 2008 show all of the characteristics we saw in the eastbound data. Running times are quite consistent except for a few days with unusual conditions described above, and there is little visible effect of peak period congestion. I have not included the charts for each segment because they are all quite similar to those for the eastbound service.
Terminal Times at Long Branch
From the segment link times, it is clear that the running time either way between Kipling and Long Branch is 5-8 minutes all day long. Therefore a round trip should take anywhere from 10-16 minutes plus whatever layover occurs at the terminal. In fact, the weekday values for this trip lie in a wide band from about 14 to 30 minutes. On weekends when the running times are at the lower end of the range, the actual round trip times show that quite long layovers were common on days when streetcar service was running through from downtown.
It can be argued that this recovery time is needed by operators who have driven all across the city, but it also shows that the scheduled trip time provides more than adequate padding. A different line management strategy and/or a shorter route would allow the same number of cars to provide more service because vehicles would not be tied up for siestas at Long Branch.
January differed from December in that the weather, while cold, did not have a single extended storm. This shows up in the generally reliable running times on the line.
What did change, however, was that a new schedule was introduced with even more running time than the previous one, and with the Long Branch service scheduled independently of the Humber service so that disruption to one branch of the line, in theory, required less intervention in service on the other one.
Running times show somewhat greater peak period effects than in much of December, but still with a fairly regular pattern. Some of the variation will be due to additional loading time with pre-Christmas loading patterns out of the way, and some to weather. Note that the rise in times inbound shows up only in the AM peak as one would expect on the outer end of a route like Queen.
Weekend times are quite flat except for Saturday, January 31, when there was an outbound delay at about 7 am.
The round trip times from Kipling to Long Branch, and by implication the terminal layovers, continue to be substantial with the highest values on New Year’s Day.
The preceding analysis shows that there is little on the western end of the Queen route to cause substantial variation in running times for vehicles. However, the actual service seen by riders is quite another matter. Inbound, the service can be affected by the accuracy, such as it is, of departure times from Long Branch Loop. Outbound, if the service is irregular leaving the congested central part of the route, it will be irregular further west, especially after half of the cars drop out at Humber. In either case, gaps caused by short turns affect service generally on the Lake Shore portion of the route.
(I will turn to the issue of service through the central-west portion of the route in a future post.)
The spread in values of headways for weekday service on page 6 of this set of charts is (a) comparable to what I have seen in previous analyses of this route and (b) totally unlike any sense of reliability one might hope for in transit service. Headways are distributed more or less evenly between the zero and 20-minute lines with several data points well above 20. The scheduled service is in the range of 11 minutes.
On the individual pages for each week, the trend lines stay near the scheduled headway, but this tends to fall apart with weather and some time-of-day effects. However, the averages mask the spread in values.
Weekend values are subject to the same problems with split route operation (buses) and weather effects as are visible in the link time charts above. Also visible is the huge spread in headways. On weekends, some values lie well above the 60 minute line, but I have not shown them to preserve a common scale for all charts.
The spread in values here is not as bad as the westbound service showing that the recovery time in the scehdule is having some benefit. However, the values overall still lie within a main band about 10 minutes wide, and many data points are under 5 or over 20 minutes. This shows that some cars are leaving very close together, and that there are some very wide gaps in onbound service. There is no time-of-day pattern to this behaviour suggesting that it is not simply a side effect of congestion and other time-specific effects elsewhere on the line.
In a future post, I will examine some specific days’ data to look at bunching problems in detail.
The data for January 2009 don’t give any comfort. Headways continued to be scattered both outbound and inbound, and if anything parts of January look worse than December even though this was supposed to be an improved schedule.
In the next article in this series, I will turn to service at the east and of the route to see how much of an effect the intensive line management deployed there is having.